Welcome

First of all thanks to SOCIS & GNU Radio for sponsoring me for these 3 months.

Histograms

I found a paper “Modulation Formats Recognition Technique Using  Artificial Neural Networks for Radio over Fiber Systems” by Guesmi and Menif, from which I originally thought it might be possible to use their histogram approach with RF, however after further experimentation with histograms of amplitudes, I’m not sure if it will be possible.

The following flow diagram represents an amplitude histogram for GFSK modulation

gfsk

While the following represents an amplitude histogram for PSK modulation

psk

The following represents an amplitude histogram for 16QAM

16qam

Azzouz & Nandi feature detection

Gamma max

I have been reading a paper from E.E.Azzouz & A.K.Nandi “Procedure for automatic recognition of analogue and digital modulations” – 1996, which presents a number of techniques to automatically classify modulation schemes of signals.

One of the classification techniques they present, is explained below:

\gamma_{max} = max |DFT(A_{cn}(i))|^2/N

A threshold is used on \gamma_{max} “to discriminate between the signals that have amplitude information and that have no amplitude information”

The amplitude of each sample passed to the DFT function, is normalized by first dividing it by the mean amplitude from the window.

The following image represents a GRC flow graph attempting to implement the above formula.

psk5.grc

I created a simple python block  which performs


print(numpy.amax(numpy.abs(in0)))

At the moment I am struggling to differentiate between QAM and GFSK etc., both modulation schemes are producing similar outputs when a channel model is utilised.

I am currently reading “Master Thesis Electrical Engineering – 2012 Automatic Modulation Recognition of Communication Signals” by Muazzam Ali Khan & Yasir Ali Bangash, who make use of the detection methods presented by Azzouz and Nandi, combining them with ANNs (Artificial Neural Networks).  I am wondering if it might be interesting to implement some of these techniques for GNU Radio making use of the Tensorflow block.

 Sigma AA

The following flow graph generates the standard deviation of normalised centred amplitudes

sigmaaa

I am currently reading Automatic Modulation Classification of communication signals – Zaihe Vu for alternative classification techniques.

Please see https://github.com/chrisruk/flowgraphs for the aforementioned flow graphs.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s